Week 1: Building Out an AI Software for LOs
This week I decided to do something slightly insane.
After talking it through with my partner Deven Gillen and loan officer Kyle Stringham, we decided to start building something new for the mortgage industry.
Not another AI chatbot.
Not another automated CRM tool.
Most CRMs already offer basic automation. That’s not the problem loan officers are trying to solve.
What we want to build is something different.
The Idea: AI That Finds the Deals
The goal is to create an AI agent that scans a loan officer’s database and finds the people most likely to need a mortgage conversation right now.
The system analyzes signals from CRM and LOS data to identify potential opportunities like:
People who may be ready to refinance
Contacts who could be preparing to buy again
Past clients who may benefit from a mortgage review
Leads who have gone quiet but may still be active
Once those opportunities are identified, the AI sends personalized outreach through text or email at the right moment.
But here’s the key difference.
The AI does not replace the loan officer.
Instead, the goal is to start the conversation and book a call with the human loan officer, keeping the relationship side of the business exactly where it belongs.
Week 1: Learning the Tech Stack
The first part of the week was spent learning the tools that actually make software run.
Each part of the system has a different role:
One tool generates the code
Another stores the code on my computer
Another tracks changes online
Another actually runs the software on the internet
If you’ve never built software before, the ecosystem alone can feel overwhelming.
And things definitely didn’t go smoothly.
The Problems We Ran Into
Like most early software projects, the first week involved a lot of troubleshooting.
Some of the issues we ran into included:
The database refusing to connect
The server failing to start
Login authentication breaking
Files being saved in the wrong locations
At several points it felt like every new step created another problem.
But each issue forced us to learn something new, and one by one we worked through them.
What We Got Working
By the end of the week, we had some real infrastructure in place.
The system now includes:
• A cloud database to store information
• A login and signup system for users
• Security protections to safeguard client data
• Separate data environments for each loan officer
This ensures every loan officer using the platform will have their own protected workspace and database.
Connecting the System to Go High Level
One of the biggest milestones this week was connecting the system to Go High Level, which many loan officers already use to manage their CRM and database.
After setting up the integration, we received our first real response from a CRM account.
It might sound small, but that moment proved the system can communicate with a live database.
That’s a major step toward making the software usable in the real world.
What Happens Next
Now that the infrastructure is in place, the next step is where things get interesting.
Next week we’ll start:
Pulling real contact data from CRM systems
Building the AI logic that identifies potential mortgage opportunities
Testing outreach workflows that can start real conversations
The goal is simple:
Help loan officers identify deals already sitting inside their database.
Because most loan officers already have opportunities in their CRM… they just don’t have a system that surfaces them at the right time.
Building in Public
We’re going to document this process week by week as we build the platform.
The wins.
The mistakes.
The lessons learned along the way.
Week 1 is officially in the books.
Building in public starts now. 🤙
